Metataxonomic sequencing reveals compositional shifts within the denture-associated microbiome in pneumonia

Background Bacterial pneumonia affects a disproportionate number of the elderly in the UK, with substantial morbidity and mortality. Mounting evidence implicates removable dentures as a potential nidus for respiratory pathogens to form a reservoir which could seed colonisation and infection of respiratory tissues in susceptible individuals. However, research evaluating the denture-associated microbiome in patients with an active diagnosis of pneumonia is lacking. The aim of this study was to characterise denture-associated oral bacterial communities by metataxonomic sequencing of 16S rRNA genes. The prevalence of antimicrobial resistance among two representative pathogenic species Staphylococcus aureus and Pseudomonas aeruginosa was also assessed. Finally, the role of salivary cytokines as diagnostic biomarkers was explored. Results There were significant shifts observed in species composition, diversity and richness in the denture-associated microbiome of pneumonia patients. Importantly, the relative abundance of putative respiratory pathogens in the denture-associated microbiota of pneumonia patients was significantly increased compared with respiratorily healthy care home residents. The magnitude of this increase was approximately three-fold in denture-associated bacterial communities compared with other oral sites examined. Antimicrobial resistance was equivocal between microbes isolated from both participant cohorts, highlighting the potential for oral biofilms to protect microbes from systemic antimicrobial therapy. While salivary cytokine profiles did not correlate with pneumonia status, the concentration of IL-6 and IL-8 positively correlated with the relative abundance of putative respiratory pathogens on denture surfaces. Conclusions This is the first study to directly examine compositional shifts in the denture-associated oral microbiome in respiratory infection, providing a basis for disentangling potential causal relationships.


Background
Lower respiratory tract infections, including pneumonia, are the fourth leading cause of death worldwide, and the most common cause of death due to infectious disease 1 .
Globally, pneumonia has a bimodal distribution of incidence, affecting the very young and elderly. However, in the United Kingdom, much of Europe and the USA, pneumonia demonstrates a predilection for the elderly, with a ten-fold increase in pneumonia cases in patients over 65 years of age 2 and 85% of pneumonia-related deaths occurring in individuals over the age of 60 years 3 .
The term pneumonia describes a clinical phenotype of acute inflammation in the lower respiratory tract 4 which does not necessarily reflect an infectious aetiology. However, the vast majority of pneumonias occur secondary to microbial infection, which may be viral, bacterial, fungal, or polymicrobial 5 . In the UK and much of Europe, pneumonia is most frequently bacterial in aetiology 6 . Diagnosis of pneumonia is challenging due to the nonspecific clinical signs and symptoms associated with the disease. Determining a microbial aetiology is confounded by the difficulty in obtaining a representative sample free from contaminating microorganisms originating in uninfected regions of the respiratory tissues or oropharynx, and the inability to distinguish microbes colonising the respiratory tissues from infective species 7 .
A burgeoning body of research has revealed an association between changes in oral microbial communities and respiratory infection in susceptible individuals 8,9,10,11 . This is most clearly supported in ventilator-associated pneumonia (VAP) which can affect mechanically ventilated intensive care patients. Here, an increase in the relative abundance of putative respiratory pathogens (PRPs) in dental plaque occurs following intubation of patients in intensive care, with a subsequent reversal of this compositional perturbation following extubation 12,13 . Further, a recent systematic review found evidence supporting the effectiveness of oral care to reduce VAP, although the effect size was modest and the overall quality of evidence available was low 14 . Similarly, a number of researchers have recovered PRPs from denture surfaces 15,16,17 , while enhanced oral care, including denture care, has been found to reduce the incidence of pneumonia among longterm care facility residents 18 . The presence of an endotracheal tube offers a direct conduit to the lungs and necessitates open mouth posture, facilitating the acquisition of exogenous microorganisms; bypassing the host immune system and enabling translocation to the respiratory tissues 19 . That a similar relationship appears to exist between the denture-associated oral microbiota and respiratory infection suggests that the presence of an artificial biomaterial surface may itself promote colonisation by PRPs, forming a reservoir that can seed infection of the respiratory tissues in susceptible individuals.
Despite indirect evidence suggesting that the oral microbial communities of denturewearing individuals may contribute to pneumonia risk, direct support for a mechanistic role for denture biomaterial surfaces in promoting respiratory infection is lacking. Therefore, this study aimed to compare the community composition of denture-associated oral bacteria in patients with a clinical diagnosis of pneumonia with respiratorily healthy care home residents. The potential role of salivary cytokines as biomarkers for both pneumonia status and oral PRP bioburden was also evaluated.

Participant demographics and clinical characteristics
Participant demographic information is summarised in Table 1. Pneumonia patients were significantly younger than care home residents (Mean difference 4 years, p = 0.0006). All pneumonia patients received antibiotic therapy (intravenous amoxicillin and clarithromycin, n = 15; other, n = 11), while only 4 of the 35 care home residents included had received antibiotics in the preceding 6 months. Otherwise there were no significant differences observed between participant cohorts. Wisdom teeth were not included in this score. + Denture Cleanliness Index scores denture cleanliness from 0 (pristine denture surfaces) to 4 (damaged dentures).

Culture isolation and antimicrobial susceptibility testing of target microorganisms
Candida was recovered from almost 90% of patients' oral cavities, an unsurprising finding given the known high prevalence of this yeast in denture-wearing individuals 29 . Both S. aureus and P. aeruginosa, two pathogens frequently associated with respiratory infection and a range of healthcare associated infections, were recovered from the oral cavities of individuals in both cohorts ( Table 2). There was no statistically significant difference between recovery rates between patients with pneumonia and respiratorily healthy individuals.

Analysis of metataxonomic sequencing data
Analysis of metataxonomic sequencing data revealed an increased relative abundance of

Analysis of salivary cytokine profiles
The level of proinflammatory cytokines assessed in saliva did not significantly differ with pneumonia status ( Figure 5). However, further analysis by linear regression revealed a significant association of IL-6 and IL-8 with the cumulative abundance of PRP species in oral (data not shown) and denture ( Figure 6) samples. The proportion of PRPs present in denture samples accounted for approximately 30% of the variability in cytokine level.

Figure 5: Salivary cytokine expression levels (pg/ml)
Note that the Y axis uses a log 10 scale.
Individual points represent each sample measured. Horizontal black line indicates median expression level for each cohort.

Discussion
While there has been mounting interest in exploring artificial biomaterial surfaces in the oral cavity as potential reservoirs of respiratory pathogens, this was the first study to directly explore compositional shifts in the denture-associated oral microbiome correlated with pneumonia status; using contemporary molecular techniques to limit selectivity bias.
Not only was an increased bioburden of putative respiratory pathogens in individuals with pneumonia found; there was a concurrent loss of species richness and diversity typically associated with a dysbiotic shift in the microbial community. Importantly, these differences were especially pronounced in denture samples, highlighting the role of dentures as a possible nidus for respiratory infection. The potential for salivary cytokines as biomarkers of pneumonia was also assessed. While cytokine levels were highly variable and did not correlate with pneumonia status directly, there were statistically significant associations of the levels of IL-6 and IL-8 with the total bioburden of respiratory pathogens in denture samples. These markers may therefore offer promise as adjuncts to identify individuals at risk of respiratory infection secondary to oropharyngeal colonisation by PRPs.
In order to reach, colonise and infect the lungs, bacteria must either pass from an external source through the oral cavity, or intrinsically through the gastrointestinal tract 30 . Thus, the relationship seen between the oral microbiome and pneumonia status may hold diagnostic potential, due to the close anatomical approximation of the oral cavity with the lungs and gastrointestinal tract, and the interface formed with the external environment. Given the poor reliability of sampling the infected lung 7 , which must be performed essentially 'blind', the ease of access to the oropharynx for microbial sampling could lead to rapid, reliable identification of potential causative microorganisms, and provide antimicrobial susceptibility profiles to aid diagnosis and treatment of pneumonia 31 .
Recruitment of eligible participants was a major challenge encountered during this study as many care home residents were cognitively impaired and thus unable to consent.
Similarly, a number of pneumonia patients had cognitive impairment either as a background comorbidity or due to acute delirium. The cross-sectional design of this study was another limitation. As recruited respiratory ward patients had received a diagnosis of pneumonia prior to recruitment, it was not possible to track changes in composition of the oral microbiota from respiratory health to disease. Similarly, there was no follow-up to examine shifts in microbial communities upon resolution of pneumonia. It was therefore not possible to determine if changes in the oral microbiome preceded pneumonia onset, a key step in determining causality 32 .
All patients with suspected pneumonia received empirical antibiotic therapy according to local policy, which reflects the British Thoracic Society guidelines on the management of severe community acquired pneumonia 33 . As only a low proportion of care home residents had received any antimicrobials in the preceding 30 days, differential antibiotic use is a potential confounder for the altered oral microbial composition seen. However, a number of factors suggest that while antibiotic use may have contributed to reduced community diversity and species richness, the differences cannot be entirely explained by antibiotic use alone. Firstly, it would be expected that denture-associated biofilms would be least affected by antibiotic use compared with other oral sites, as biofilms may confer antimicrobial tolerance to constituent microbes 34 . Moreover, antibiotics have to traverse the oral mucosal barrier, diffuse through the palatal microbial biofilm and then penetrate the denture-associated biofilm in sufficient concentration to perturb microbial communities.
It should be noted that Enterobacteriaceae are typically not susceptible to macrolide antibiotics such as clarithromycin and are intrinsically resistant to amoxicillin and other beta-lactamases 35 . The aggressive use of these antibiotic regimes in pneumonia patients may act as a selective pressure to suppress growth and survival of normal oral microbes, particularly Streptococcaceae, leading to an increased relative abundance of more virulent microorganisms 36 . Nonetheless, the finding that the difference in relative abundance of PRPs between cohorts was most pronounced in denture samples suggests that antibiotic use was unlikely to be the primary contributor to the changes in microbial community composition. Notably, no S. aureus isolates recovered from respiratory ward patients were resistant to amoxicillin, compared to over one quarter of those from care home residents.
However, macrolide resistance was more than doubled in respiratory ward S. aureus isolates. There were much higher rates of resistance to the beta-lactam antibiotic piperacillin-tazobactam in P. aeruginosa isolates from care home residents compared with pneumonia patients, as was seen for the related cephalosporin ceftazidime. However, resistance of P. aeruginosa isolates to ciprofloxacin, a fluoroquinolone antibiotic, was found to be much higher among pneumonia patients than care home residents. While the low number of both S. aureus and P. aeruginosa isolates recovered precludes any reliable statistical evaluation, the equivocal resistance patterns observed suggest that antibiotic treatment may not have exerted a major selective pressure upon the oral microbiota. This was particularly evident in the case of S. aureus isolates, where amoxicillin sensitive strains were isolated from respiratory ward patients' samples despite empiric therapy with this agent.

Conclusions
This study revealed that perturbations within the denture-associated oral microbiome are

Participant recruitment and sample collection
We recruited a total of 66 denture-wearing individuals from long term residential care facilities (n = 35) and hospital wards (n = 26). Participants were excluded from either group if they lacked capacity to provide consent; were receiving palliative end- isolates was tested according to the EUCAST disc-diffusion method 22 .
Salivettes were centrifuged twice at 3000 x g to recover decellularised saliva, from which a panel of pro-inflammatory cytokines were subsequently analysed using a Cytometric Bead Array kit (BD Biosciences, Wokingham, UK) and flow-assisted cell sorting (BD FACSCanto II, BD Biosciences, Wokingham, UK). Cytokine profiles were assessed according to pneumonia status and relative abundance of denture-associated potential respiratory pathogens.

Streptococcus pneumoniae by qPCR and 16S rRNA gene sequencing
Microbial swabs were aseptically transferred to 10 ml bijou bottles containing 1 ml of 0.9% Rare operational taxonomic units OTUs (<10 reads) were excluded from further analysis and any OTUs with less than 98% coverage or 97% sequence identity to a known bacterial species were categorised to genus level only. After manual scanning, OTUs that would not be expected to occur in the oral cavity were re-examined using the NCBI BLASTn database.

Statistical analyses
Statistical analysis was conducted using R 25

Ethics declarations
Ethical approval for this study was obtained from the Wales REC 6; reference 16/WA/0317.
All participants provided written consent for this study.

Consent for publication
Not applicable      Linear regression of salivary cytokine level against cumulative relative abundance of PRP species 95% confidence intervals indicated by dashed lines. P values represent significance following F test with Bonferroni post-hoc correction.